Re: [PATCH v1 17/17] drm/mediatek: Add mt8195-dpi support to drm_drv

[Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>]

On 05/10/2022 11:34, Guillaume Ranquet wrote:
> On Tue, 04 Oct 2022 17:05, Krzysztof Kozlowski
> <krzysztof.kozlowski@linaro.org> wrote:
>> On 04/10/2022 13:55, Guillaume Ranquet wrote:
>>>> No. You said what the code is doing. I think I understand this. You
>>>> still do not need more compatibles. Your sentence did not clarify it
>>>> because it did not answer at all to question "why". Why do you need it?
>>>>
>>>> Sorry, the change looks not correct.
>>>>
>>>> Best regards,
>>>> Krzysztof
>>>>
>>>
>>> I need a new compatible to adress the specifics of mt8195 in the mtk_dpi driver,
>>> the change is in this series with:
>>> [PATCH v1 16/17] drm/mediatek: dpi: Add mt8195 hdmi to DPI driver [1]
>>
>> But you do not have specifics of mt8195. I wrote it in the beginning.
>>
>>>
>>> I then need to add that compatible to the "list" here in mtk_drm_drv.
>>
>> No, you do not... I checked the driver and there is no single need... or
>> convince me you need.
>>
>>> I don't see a way around this unless I rewrite the way mtk_drm_drv works?
>>
>> Why rewrite? You have all compatibles in place.
>>
>>>
>>> Maybe if I declare a new compatible that is generic to all mediatek
>>> dpi variants?
>>
>> You were asked to use fallback. Don't create some fake fallbacks. Use
>> existing ones.
>>
>>> and have all the dts specify the node with both the generic dpi and
>>> the specific compatible?
>>>
>>> dpi@xxx {
>>> 	compatible = "mediatek,dpi", "mediatek,mt8195-dpi";
>>
>> I don't know what's this but certainly looks odd. Some wild-card
>> compatible in front (not fallback) and none are documented.
>>
>>> 	...
>>> }
>>>
>>> Then I can "collapse" all the dpi related nodes in mtk_drm_drv under
>>> "mediatek,dpi" ?
>>>
>>> I guess would have to do the change for all other components that are needed in
>>> mtk_drm_drv (mmsys, aal, ccor, color, dither, dsc, gamma, mutex...).
>>>
>>> That's the only trivial way I can think of implementing this with the
>>> current status
>>> of the mtk_drm stack.
>>>
>>> Do you have any other ideas in mind?
>>
>> Use fallback of compatible device. That's the common pattern.
>> Everywhere, Mediatek as well.
>>
> 
> I'm unsure about what a "fallback of compatible device" is.
> But from what I can gather, you'd have me write the dts as:
> 
> dpi@xxx {
> 	compatible = "mediatek,mt8195-dpi", "mediatek,mt2701-dpi";
> 	...
> }

Sounds reasonable. Just someone should check whether devices are
actually compatible, because driver is just a hint.

> 
> so that the mtk_dpi driver will use the specific mt8195 dpi config and
> the mtk_drm driver will fallback to mt2701 to find the compatible it needs.
> 
> Would you like me to use this pattern for all the other compatibles declared
> in the mtk_ddp_comp_ids array in separate patches?

Could be for consistency, but it is up to you.

Best regards,
Krzysztof